Geospatial Analysis of MRT Connectivity Gaps and a New Transit Proposal for Singapore

Part 2

Singapore MRT Morning Rush Hour Analysis

This map visualizes the Morning Out-In Passenger Ratio across the MRT network in Singapore during weekday peak hours. The sample of the data is taken from the weekdays in April 2025. Each circle represents an MRT station, with its size indicating passenger volume and color reflecting the Out-In Ratio:

• Yellow represents stations with balanced or inbound-heavy flows (more people entering than exiting).
• Orange to dark red represents stations with a high out-in ratio (significantly more people exiting than entering), typical of commuter destination

Key Observations:

• Downtown Core and CBD stations (e.g., Raffles Place, City Hall, Tanjong Pagar) exhibit dark red tones and large sizes, indicating they are major destinations with high volumes of alighting passengers in the morning.

• Other stations with a high out-in ratio includes One-North, Labrador Park, Expo, Kent Ridge, Tai Seng and stations in Tuas.

• Residential towns such as Jurong East, Woodlands, Pasir Ris, and Punggol show lighter colors and large circles, reflecting high numbers of boarding commuters starting their daily journeys.

• The flow pattern suggests a strong directional movement from peripheral residential areas into the central business district during the morning rush.

This map provides insights into urban land use, indicating the concentration of employment hubs versus residential catchments.

Singapore MRT Morning Rush Hour Average Number of Tap Ins

This metric represents the average number of passengers entering (tapping into) each MRT station during the morning peak hours, typically defined as 7AM to 9AM on weekdays. It serves as a key indicator of commuter origin intensity, revealing where large numbers of people begin their daily transit journeys.

The plot shows that stations in residential areas like Punggol, Tampines, Jurong, Woodlands, Ang Mo Kio etc have higher average number of tap ins in comparison with stations in the CBD area.

Singapore MRT Morning Rush Hour Average Number of Tap Outs

This metric is a strong indicator of commuter destination hotspots, particularly for workplaces, schools, and business hubs. This reinforces the idea that Stations in the Central Business District (CBD) — like Raffles Place, City Hall, Tanjong Pagar, Bugis, Dhoby Ghaut, and Orchard — tend to have high average tap-out counts in the morning. These areas contain offices, government buildings, universities, and shopping centers, which are key destination drivers.

Distance to Nearest MRT by Subzones

This map shows the average distance from each subzone in Singapore to the nearest MRT station, useful for understanding transit accessibility and urban connectivity.

Map Explanation:

• A choropleth map of Singapore, where each polygon represents a URA planning subzone.
• MRT lines are overlaid in colored routes (e.g., North-South Line in red, East-West Line in green, etc.).
• Color represents the average distance (in kilometers) from locations within each subzone to the nearest MRT station.
  • Lighter yellow: Subzones with excellent MRT access, typically < 0.5 km.
  • Darker red: Subzones where average distance is > 2 km, indicating poorer MRT access.

This map illustrates the average distance from each subzone in Singapore to the nearest MRT station. While central and mature residential towns have excellent access within 500 meters, several peripheral subzones, particularly in the west and northeast—remain over 2 km away from MRT stations. This insight can inform transit planning, identify underserved regions, and guide future land development policies aligned with accessibility goals.

Map Of Underserved Subzones

A more composite approach which takes into account population density, distance to nearest mrt station and ratio of mrt users of a subzone is used to identify subzones with potential MRT access gaps.

Methodology
I evaluated each subzone based on the following three dimensions:

1. Proximity to MRT:

   • Subzones where the average distance to the nearest MRT station exceeds 0.8 km were considered less accessible.

2. Population Density:

   • Subzones with above-median population density were considered to have sufficient residential concentration to justify MRT demand.

3. MRT Usage Ratio:

   • The proportion of residents using MRT as a transport mode was used to infer actual demand.

Subzones were categorized using the 30th and 70th percentiles of Ratio_of_MRT_Users:

• Low usage (< 30th percentile)
• High usage (> 70th percentile)

Classification of MRT Access Gaps
Two distinct MRT access gap types were flagged:

1. Suppressed Demand Zones
   Subzones where MRT usage is low despite high population density and poor proximity to MRT.
   Interpretation:
   These areas might reflect latent or unmet demand — residents could use MRT more if access were improved (e.g., new stations, better feeder connectivity).

2. Overloaded Demand Zones
   Subzones with high MRT usage, high density, but still located far from MRT stations.
   Interpretation:
   These areas are potentially over-reliant on limited MRT access points, possibly leading to station or train overcrowding.

Flagging MRT Access Gap Areas I created a composite boolean flag MRT_Access_Gap, which is set to True if either of the above conditions is met. This identifies subzones that:

• Are densely populated,
• Are located far from an MRT station,
• And either exhibit suppressed or overloaded usage patterns.

Implications on Policy Making / Urban Planning
Subzones identified as having MRT access gaps can be:

• Prioritized for feeder bus enhancements, last-mile solutions, or future station planning.
• Considered in Transit-Oriented Development (TOD) frameworks.

Identified Underserved Subzones:

1. Dover
2. Balestier
3. Tai Seng
4. Frankel
5. Trafalgar
6. Pasir Rist West
7. Flora Drive
8. Yishun East
9. Northland
10. Sembawang North

K-Means Clustering

To explore hidden patterns in how subzones relate to MRT accessibility and demand, we applied K-Means clustering.

Objective
Cluster subzones into distinct categories based on:

• Physical access to MRT,
• Population density (residential and working),
• MRT usage patterns

Steps in Preparing Features
cols_to_use = ['dist_to_nearest_mrt_km', 'population_density', 'Working Population Density', 'Ratio_of_MRT_Users']
These variables represent key dimensions of MRT access and potential demand.

• dist_to_nearest_mrt_km: How far the average person is from an MRT station.
• population_density & Working Population Density: Size of the residential and commuter populations.
• Ratio_of_MRT_Users: Actual reliance on MRT.

an inverted MRT usage metric is created gdf_cluster['mrt_user_inverse'] = 1 - gdf_cluster['Ratio_of_MRT_Users'] This lets the algorithm treat lower MRT usage as an indication of potential access issues or latent demand, aligning better with the clustering intent.

Since the variables are on different scales (e.g., kilometers vs. percentages vs. population per km²), standardization (z-score normalization) is needed. StandardScaler()
This ensures that no single variable dominates the clustering.

The algorithm groups subzones into 4 clusters, where each cluster shares similar patterns across the selected features. I interpret each cluster as a type of MRT access-demand profile:
X-axis: Distance to Nearest MRT (km)

Y-axis: Population Density (per km²)

Z-axis: Working Population Density (per km²)

Color: Cluster assignment (0 = red, 1 = blue, 2 = yellow, 3 = green)

Here’s a breakdown of each cluster's characteristics based on their location in the plot:

Cluster 0
High Residential + Working Density, Closer to MRT

• Distance to MRT: Mostly near MRT stations (left side of x-axis).
• Pop & Work Density: Moderate to high.
  Interpretation:
  These are central or mature urban zones with high demand and decent MRT coverage.
  Possible Actions:
  Monitor crowding and plan capacity improvements.
  These zones may already be well served, but demand could exceed supply.

Cluster 1
High Distance from MRT, Medium Density

• Distance to MRT: Largest among all clusters (right side of x-axis).
• Density: Low to medium for both residents and workers.
  Interpretation:
  Underserved or outlying areas—people live or work far from MRT, yet the population is not sparse.
  Possible Actions:
  Potential candidates for new MRT stations, feeder bus enhancements, or first-mile last-mile solutions.
  Good focus for equity-driven planning.

Cluster 2
Extremely High Working Density, Near MRT

• Distance to MRT: Very close (leftmost).
• Working Pop: Extremely high (top of z-axis).
  Interpretation:
  These are likely commercial or CBD zones with heavy commuter traffic.
  Possible Actions:
  Maintain strong MRT connectivity.
  May need crowd control strategies, e.g., staggered working hours or transit signal upgrades.

Cluster 3
Low Density, Moderate Distance

• Distance to MRT: Mid-range.
• Population & Working Density: Medium Low.
  Interpretation:
  These are low-density subzones that are relatively far from MRT stations, suggesting lower accessibility and possibly underserved regions in terms of transport.
  Possible Actions:
  Improve First-Last Mile Connectivity Future MRT/LRT Expansion

Cluster	MRT Distance	Pop Density	Work Density	Interpretation	Action Suggestion
0 (Red)	Low	High	High	Urban centers, well-served	Capacity improvement
1 (Blue)	High	Medium	Medium	Underserved Mid-density zones	MRT access expansion
2 (Yellow)	Very Low	Medium	Very High	CBD/commercial hubs	Transit demand management
3 (Green)	Medium High	Medium Low	Medium Low	Suburbs / Industrial Area, Possibly underserved	MRT access expansion

Segmenting subzones allows planners to prioritize interventions based on cluster profiles.

Each group can receive targeted strategies:

1. Infrastructure investment
2. Encouraging use of different transport methods
3. Feeder service improvements

Means of Different Metrics by Cluster Groups

	dist_to_nearest_mrt_km	population_density	Working Population Density	Ratio_of_MRT_Users
cluster
0	0.460729	6935.100353	9117.760471	0.520259
1	0.467292	33420.074271	3575.596250	0.362344
2	0.201250	1175.575625	59607.884375	0.681437
3	1.129031	4491.511959	2391.653608	0.325351

Clustering Group of Underserved Subzones

Each of the underserved subzones identified above is further categorized into different clustering groups so different and more precise approach can be prescribed on each underserved subzones.

         SUBZONE_N cluster              Cluster Zone Type
0            DOVER       0     Urban Centers, Well-served
1          FRANKEL       3         Suburbs / Sparse Areas
2        BALESTIER       0     Urban Centers, Well-served
3      FLORA DRIVE       3         Suburbs / Sparse Areas
4         TAI SENG       3         Suburbs / Sparse Areas
5   PASIR RIS WEST       3         Suburbs / Sparse Areas
6        TRAFALGAR       1  Underserved Mid-Density Zones
7      YISHUN EAST       1  Underserved Mid-Density Zones
8  SEMBAWANG NORTH       3         Suburbs / Sparse Areas
9        NORTHLAND       3         Suburbs / Sparse Areas

Proposing Solutions

For underserved subzones in cluster 0 (Dover & Balestier) where both residential and working population is high and the distance to MRT Station is small, the solution is to improve the capacity of the nearest mrt station to the subzones or add more feeder bus services between the subzone and the nearest MRT stations.

Underserved cluster 1 subzones (Yishun East & Trafalgar), despite having moderate population and working densities, are located far from MRT stations, indicating potential desperately underserved areas. The long distance to the nearest MRT Stations can be addressed by improving connectivity through adding feeder bus services, improve pedestrian walkway etc. These subzones are also ideal for future construction of MRT stations.

There is no underserved cluster 2 subzones as all cluster 2 subzones are in the CBD area and are already well served by MRT service.

For undersevred subzones in cluster 3 (Frankel, Tai Seng, Flora Drive, Pasir Ris West, Northland & Sembawang North). These are relatively low-density subzones that are relatively far from MRT stations, suggesting lower accessibility and possibly underserved regions in terms of transport. First solution is improve first-last mile connectivity by providing more feeder buses and Increase frequency and route coverage for these areas. These subzones should also be prioritized for future MRT/LRT expansion plans. Light rail or automated shuttles can be considered for smaller population centers

Proposed MRT Line: East Coast–North East Line (ENL)

The ENL is a proposed MRT line that aims to connect underserved subzones in Singapore’s eastern and northeastern regions. It spans from Marine Terrace in the East Coast to Yishun, cutting across key areas that currently face poor MRT accessibility and are part of the cluster 1 (mid density, high MRT distance) and cluster 3 (low density, high MRT distance) in the map.

Stations:

EN1: Marine Terrace
EN2: Frankel
EN3: Kembangan (connects to EWL)
EN4: Kaki Bukit
EN5: Paya Lebar Air Base (redevelopment site)
EN6: Defu (connects to CRL)
EN7: Hougang (connects to NEL, CRL)
EN8: Trafalgar
EN9: Seletar Hills
EN10: Sengkang West (near emerging residential zones)
EN11: Seletar Aerospace Park (employment hub)
EN12: Yishun East
EN13: Yishun (connects to NSL)

Benefits of New MRT Line to the Region:

1. Improves Access for Underserved Areas
   Links low-access regions like Frankel, Trafalgar, Seletar, and Yishun East directly to the MRT network.

2. Supports Transit-Oriented Redevelopment
   Paya Lebar Air Base will undergo massive redevelopment to become new residential areas. ENL supports future residential and commercial growth here.
   Encourages high-density, car-lite developments along the corridor.

3. Enhances First-Last Mile Connectivity
   Reduces reliance on feeder buses and long commutes for residents in Seletar, fringes of Hougang, and Frankel.
   Makes existing bus services more efficient with MRT as a backbone.

4. Boosts Employment Accessibility
   Seletar Aerospace Park becomes directly accessible via MRT, giving job access to residents from East Coast and Northeast.
   Connects residential neighborhoods to industrial/commercial areas.

5. Relieves Congestion on Existing Lines
   Offers an alternative north-south corridor, easing load on NEL, EWL, and NSL.
   Diverts intra-region travel that currently congests core interchange stations.

6. Provide More Interchange Opportunities, Higher Network Integration & Supports Decentralization Goals ENL enhances cross-island connectivity by providing interchange opportunities with multiple major MRT lines.
   These interchanges include:

Hougang (EN7) → Connects with the North East Line (NEL)
Yishun East (EN12) → Connects with the North South Line (NSL)
Kaki Bukit (EN4) → Connects with the Downtown Line (DTL)
Kembangan (EN3) → Connects with the East West Line (EWL)
Marine Terrace (EN1) → Connects with the Thomson-East Coast Line (TEL)

These connections allow commuters to transfer between high-capacity lines more efficiently, facilitating smoother travel across Singapore without needing to detour through the central interchange stations. This relieves congestion, reduces travel time, and supports the development of a more resilient, decentralized MRT network.

[NbConvertApp] Converting notebook Singapore_MRT_Analysis.ipynb to html
C:\Users\Chia Hon Lam\AppData\Local\Programs\Python\Python310\lib\site-packages\zmq\_future.py:718: RuntimeWarning: Proactor event loop does not implement add_reader family of methods required for zmq. Registering an additional selector thread for add_reader support via tornado. Use `asyncio.set_event_loop_policy(WindowsSelectorEventLoopPolicy())` to avoid this warning.
  self._get_loop()
[NbConvertApp] Writing 19961431 bytes to Singapore_MRT_Analysis.html